the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
PMF-LP: the first 10 m plastic-mulched farmland distribution map (2019–2021) in the Loess Plateau of China generated using training sample generation and classifier transfer method
Abstract. Plastic film mulching has been extensively used to increase crop yields in arid and semi-arid regions, but it also altered agricultural landscapes and caused severe environmental pollution. Therefore, accurate and timely mapping of plastic-mulched farmland (PMF) distributions is crucial for planning agricultural production and preventing micro-plastic pollution. However, the scarcity of sufficient and representative training samples hinders large-scale supervised classification and extraction of PMF. Additionally, it remains unclear whether a pre-trained classifier can be directly applied to different regions and years for rapid PMF mapping. To address these challenges, we proposed a new framework that simultaneously takes advantages of the automation of index-based method and the generalization ability of supervised classifier-based approach for PMF mapping. Based on the distinctive spectral responses induced by plastic film deployment events, two novel and robust PMF indices—the Max Blue Band-based Plastic-mulched Farmland Index (MBPMFI) and the Blue Band-based Plastic-mulched Farmland Index (BPMFI)—were initially designed to automatically and rapidly extract PMF pixels in cloud-free areas as candidate training samples. Additionally, the transferability of classifiers pre-trained with these automatically generated samples and optimal features was further evaluated in spatial and spatial–temporal transferability scenarios using F1 values. Finally, by coupling the index-based training sample generation method with the temporal classifier transfer approach, PMF distributions were rapidly produced for the Loess Plateau of China (PMF-LP) for 2019–2021. The results showed that the two newly established indices, MBPMFI and BPMFI, were more robust than the existing PMF indices in enhancing PMF information and suppressing complicated backgrounds. The temporal classifier transfer proved suitable for directly and rapidly mapping PMF across multiple years without additional training samples. Using the locally adaptive classifiers as a reference, the average accuracy decrease of the transferred classifiers was less than 7.0 % under the temporal transferability scenario. Our mapping framework achieved F1 values of 0.80–0.86 in recognizing PMF distributions for the Loess Plateau, highlighting its ability to delineate large-scale spatial patterns of PMF. Additionally, the estimated PMF areas based on the PMF-LP aligned well with the agricultural census data at municipal level (R2 > 0.87). The framework developed in this study lays a foundation for future monitoring of PMF distributions and agricultural micro-plastic pollution on a large scale. The full archive of PMF-LP is freely available at https://doi.org/10.5281/zenodo.13369426 (Zhao et al., 2024).
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(21972 KB)
-
Supplement
(5052 KB)
-
This preprint has been withdrawn.
- Preprint
(21972 KB) - Metadata XML
-
Supplement
(5052 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
CC1: 'Comment on essd-2024-372', Jie Bai, 30 Oct 2024
This manuscript proposed a new method with two PMF indices (MBPMFI and BPMFI)) to automatically and rapidly extract the plastic-mulched farmland distribution in the Loess Plateau of China from 2019 to 2021. These two PMF indices were being proved to be more robust than other existing PMF indices. The accuracy of the plastic-mulching area extracted in this study can meet the needs of users by comparing with the ground points and statistical yearbook data. This work displayed the large-scale spatial patterns of PMF and calculated the mean value of recognizing PMF maps for the study areas was 0.8-0.86. This work can also be transferred to the extraction of the large plastic-mulched farmland in northwest arid regions of China. On the whole, the method of this manuscript is feasible, the results are reliable, and it meets the research field and paper publication requirements of this journal.
However, there are the following points that need to be clarified and supplemented in detail.
(1) When the author extracted the plastic-mulched farmland, he used the comparison of crop phenological curve and plastic-mulch phenological curve. Are the phenological curves of these three different crops (maize, potato, and winter wheat) in the study area very different? Does it affect the accuracy of extracting the film distribution? The author can add clarification in the discussion section.
(2) The study area is relatively large, spanning seven provinces. Are the colors of plastic mulches the same? In many farmlands of China, the color of the mulch will be different, such as transparent, white or black. The author needs to explain the color of the mulch. Is the phenological curve different for different colors of plastic mulch?
Citation: https://doi.org/10.5194/essd-2024-372-CC1 -
RC1: 'Comment on essd-2024-372', Anonymous Referee #1, 26 Jan 2025
This study tried to generate PMF samples through spectral indices and then obtain the PMF maps on the Loess Plateau of China for 2019 to 2021 using random forest. However, the data quality is unsatisfactory after careful evaluation while the study area is too small. The method to automatically generate the PMF samples has certain limitations while more samples from in-situ field surveys is a must to increase the classification performance and to justify the reliability of the PMF data product.
As for data quality, there exist large false positives as bare land and impervious surfaces like airport are misclassified into PMF while the salt-and-pepper effect is not tackled. The temporal consistency among the three years is also an issue. More specific comments and the data quality check details are in the PDF uploaded.
-
RC2: 'Comment on essd-2024-372', Anonymous Referee #2, 27 Jan 2025
General comment:
I believe the paper could benefit from aligning more closely with the ESSD manuscript type, aim, and scope. This might enhance its relevance and engagement with the audience. Although the study includes a dataset, the preprint is shown in a very long research article format.
Additional comments:
- To improve dataset reliability, field visits and visual inspections should be better described.
- The paper is too long. The title, abstract, and main text are lengthy. The entire preprint article file is 40 pages, and as described in the journal presentation quality section, long articles are not expected. All the sections could be summarized to facilitate the reading and better transmit the main contributions of the dataset.
- The authors must be congratulated on the high-quality figures in the paper. They incorporate appropriate cartographic elements as needed, using effective color combinations, sizes, and layouts.
- I believe plastic films are mainly made of polyethylene (PE), not PVC, as the authors indicated in the introduction.
- Using cropland and cloud masks limits the dataset's reuse. Information from cloudy periods could enhance machine learning (ML). Data reliability is significant for ML training focusing on PMF mapping in large regions. As correctly noted and indicated by the authors in the introduction section, the time window for PMF recognition is very short, so accurate spatial and temporal information is relevant for ML PMF mapping. Both masks can be used in a research article to improve the current information about PMF distribution, but I would use less of this technique in a data article.
- I am unclear on the method of cropland masking. I did not understand if they used the previous datasets to exclude only the coincident non-cropland pixels from both studies. The dataset must not rely on crop masks of 30m spatial resolution because some misclassification could have happened in the previous classification process, incorporating noise in the dataset. Also, applying a mask from 2020 for all the years doesn’t seem like a good practice. Is there any limitation on using the yearly mask for each respective year?
- Certain sections resemble a typical research article rather than a data description paper. I believe the authors have enough material to produce at least two articles, an original research article (the “two novel and robust PMF indices” highlighted may be a sufficient contribution), a data description article, and maybe more.
- In the Figure 6a, the image seems oversaturated in some parts.
- Figure 8 has significant issues. In the legend, the red color is associated with PMF, but this is inverted in panels 8b and 8e. The indices should be normalized to enable proper comparison, as they are currently on completely different scales. While one might argue that the white color indicates the threshold, the authors should clearly state this in the legend if this is the case.
- The issues in comment 8 are also in Figures 9 and 10.
- There is a broken link in the supplement.
- The concise metadata may require further refinement to enhance its clarity and effectiveness.
Citation: https://doi.org/10.5194/essd-2024-372-RC2
Interactive discussion
Status: closed
-
CC1: 'Comment on essd-2024-372', Jie Bai, 30 Oct 2024
This manuscript proposed a new method with two PMF indices (MBPMFI and BPMFI)) to automatically and rapidly extract the plastic-mulched farmland distribution in the Loess Plateau of China from 2019 to 2021. These two PMF indices were being proved to be more robust than other existing PMF indices. The accuracy of the plastic-mulching area extracted in this study can meet the needs of users by comparing with the ground points and statistical yearbook data. This work displayed the large-scale spatial patterns of PMF and calculated the mean value of recognizing PMF maps for the study areas was 0.8-0.86. This work can also be transferred to the extraction of the large plastic-mulched farmland in northwest arid regions of China. On the whole, the method of this manuscript is feasible, the results are reliable, and it meets the research field and paper publication requirements of this journal.
However, there are the following points that need to be clarified and supplemented in detail.
(1) When the author extracted the plastic-mulched farmland, he used the comparison of crop phenological curve and plastic-mulch phenological curve. Are the phenological curves of these three different crops (maize, potato, and winter wheat) in the study area very different? Does it affect the accuracy of extracting the film distribution? The author can add clarification in the discussion section.
(2) The study area is relatively large, spanning seven provinces. Are the colors of plastic mulches the same? In many farmlands of China, the color of the mulch will be different, such as transparent, white or black. The author needs to explain the color of the mulch. Is the phenological curve different for different colors of plastic mulch?
Citation: https://doi.org/10.5194/essd-2024-372-CC1 -
RC1: 'Comment on essd-2024-372', Anonymous Referee #1, 26 Jan 2025
This study tried to generate PMF samples through spectral indices and then obtain the PMF maps on the Loess Plateau of China for 2019 to 2021 using random forest. However, the data quality is unsatisfactory after careful evaluation while the study area is too small. The method to automatically generate the PMF samples has certain limitations while more samples from in-situ field surveys is a must to increase the classification performance and to justify the reliability of the PMF data product.
As for data quality, there exist large false positives as bare land and impervious surfaces like airport are misclassified into PMF while the salt-and-pepper effect is not tackled. The temporal consistency among the three years is also an issue. More specific comments and the data quality check details are in the PDF uploaded.
-
RC2: 'Comment on essd-2024-372', Anonymous Referee #2, 27 Jan 2025
General comment:
I believe the paper could benefit from aligning more closely with the ESSD manuscript type, aim, and scope. This might enhance its relevance and engagement with the audience. Although the study includes a dataset, the preprint is shown in a very long research article format.
Additional comments:
- To improve dataset reliability, field visits and visual inspections should be better described.
- The paper is too long. The title, abstract, and main text are lengthy. The entire preprint article file is 40 pages, and as described in the journal presentation quality section, long articles are not expected. All the sections could be summarized to facilitate the reading and better transmit the main contributions of the dataset.
- The authors must be congratulated on the high-quality figures in the paper. They incorporate appropriate cartographic elements as needed, using effective color combinations, sizes, and layouts.
- I believe plastic films are mainly made of polyethylene (PE), not PVC, as the authors indicated in the introduction.
- Using cropland and cloud masks limits the dataset's reuse. Information from cloudy periods could enhance machine learning (ML). Data reliability is significant for ML training focusing on PMF mapping in large regions. As correctly noted and indicated by the authors in the introduction section, the time window for PMF recognition is very short, so accurate spatial and temporal information is relevant for ML PMF mapping. Both masks can be used in a research article to improve the current information about PMF distribution, but I would use less of this technique in a data article.
- I am unclear on the method of cropland masking. I did not understand if they used the previous datasets to exclude only the coincident non-cropland pixels from both studies. The dataset must not rely on crop masks of 30m spatial resolution because some misclassification could have happened in the previous classification process, incorporating noise in the dataset. Also, applying a mask from 2020 for all the years doesn’t seem like a good practice. Is there any limitation on using the yearly mask for each respective year?
- Certain sections resemble a typical research article rather than a data description paper. I believe the authors have enough material to produce at least two articles, an original research article (the “two novel and robust PMF indices” highlighted may be a sufficient contribution), a data description article, and maybe more.
- In the Figure 6a, the image seems oversaturated in some parts.
- Figure 8 has significant issues. In the legend, the red color is associated with PMF, but this is inverted in panels 8b and 8e. The indices should be normalized to enable proper comparison, as they are currently on completely different scales. While one might argue that the white color indicates the threshold, the authors should clearly state this in the legend if this is the case.
- The issues in comment 8 are also in Figures 9 and 10.
- There is a broken link in the supplement.
- The concise metadata may require further refinement to enhance its clarity and effectiveness.
Citation: https://doi.org/10.5194/essd-2024-372-RC2
Data sets
PMF-LP: the first 10 m plastic-mulched farmland distribution map (2019-2021) in the Loess Plateau of China generated using training sample generation and classifier transfer method Cheng Zhao et al. https://doi.org/10.5281/zenodo.13369426
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
523 | 131 | 164 | 818 | 69 | 17 | 19 |
- HTML: 523
- PDF: 131
- XML: 164
- Total: 818
- Supplement: 69
- BibTeX: 17
- EndNote: 19
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Cheng Zhao
Yadong Luo
Xiangyu Chen
Linsen Wu
Zhao Wang
Hao Feng
Qiang Yu
Jianqiang He
This preprint has been withdrawn.
- Preprint
(21972 KB) - Metadata XML
-
Supplement
(5052 KB) - BibTeX
- EndNote